A deep learning model for personalized intra-arterial therapy planning in unresectable hepatocellular carcinoma: a multicenter retrospective study

Abstract

Background: Unresectable Hepatocellular Carcinoma (uHCC) poses a substantial global health challenge, demanding innovative prognostic and therapeutic planning tools for improved patient management. The predominant treatment strategies include Transarterial chemoembolization (TACE) and hepatic arterial infusion chemotherapy (HAIC).

Methods: Between January 2014 and November 2021, a total of 1725 uHCC patients [mean age, 52.8 ± 11.5 years; 1529 males] received preoperative CECT scans and were eligible for TACE or HAIC. Patients were assigned to one of the four cohorts according to their treatment, four transformer models (SELECTION) were trained and validated on each cohort; AUC was used to determine the prognostic performance of the trained models. Patients were stratified into high and low-risk groups based on the survival scores computed by SELECTION. The proposed AI-based treatment decision model (ATOM) utilizes survival scores to further inform final therapeutic recommendation.

Findings: In this study, the training and validation sets included 1448 patients, with an additional 277 patients allocated to the external validation sets. The SELECTION model outperformed both clinical models and the ResNet approach in terms of AUC. Specifically, SELECTION-TACE and SELECTION-HAIC achieved AUCs of 0.761 (95% CI, 0.693-0.820) and 0.805 (95% CI, 0.707-0.881) respectively, in predicting ORR in their external validation cohorts. In predicting OS, SELECTION-TC and SELECTION-HC demonstrated AUCs of 0.736 (95% CI, 0.608-0.841) and 0.748 (95% CI, 0.599-0.865) respectively, in their external validation sets. SELECTION-derived survival scores effectively stratified patients into high and low-risk groups, showing significant differences in survival probabilities (P < 0.05 across all four cohorts). Additionally, the concordance between ATOM and clinician recommendations was associated with significantly higher response/survival rates in cases of agreement, particularly within the TACE, HAIC, and TC cohorts in the external validation sets (P < 0.05).

Interpretation: ATOM was proposed based on SELECTION-derived survival scores, emerges as a promising tool to inform the selection among different intra-arterial interventional therapy techniques.

Funding: This study received funding from the Beijing Municipal Natural Science Foundation, China (Z190024); the Key Program of the National Natural Science Foundation of China, China (81930119); The Science and Technology Planning Program of Beijing Municipal Science & Technology Commission and Administrative Commission of Zhongguancun Science Park, China (Z231100004823012); Tsinghua University Initiative Scientific Research Program of Precision Medicine, China (10001020108); and Institute for Intelligent Healthcare, Tsinghua University, China (041531001).

Keywords: Artificial intelligence; Decision support; Deep-learning; Hepatocellular carcinoma.

Fig. 1

Patient Enrolment of uHCC patients. (a) Cohort who received TACE or HAIC therapy. (b) Cohort who received TACE or HAIC in combination with systemic therapies, cohort B is considered a subgroup from cohort A. Abbreviations: uHCC, unresectable hepatocellular carcinoma; TACE, Transarterial Chemoembolization; HAIC, Hepatic-arterial infusion chemotherapy.

Fig. 2

The workflow of this study, from deep learning model building, decision support pipeline construction to model evaluation. (a) Architecture of SELECTION, Multi-modal input construction was considered from clinical variables and CECT images; (b) Simplified illustration of ATOM in clinical application, where ATOM consist of four SELECTION models trained by four distinctive cohorts, each model will provide a survival score for the patient, through designed decision support pipeline, provide a treatment recommendation according to corresponding SELECTION model. (c) Attention operations in the bidirectional multimodal attention layer. (d) Evaluation methods used to assess the model performance. Abbreviations: uHCC, unresectable hepatocellular carcinoma; MLP, Multilayer Perceptron; Mul., Multiplication; SELECTION, unreSEctabLe hEpatocellular Carcinoma mulTImOdal transformer; AUROC, area under the receiver operating characteristic curve.

Fig. 3

(a) Overview of ATOM, providing uHCC patient with therapeutic decision planning recommendation. (b) Patient management flowchart demonstrating the risk stratification process, and the ATOM recommendation of therapeutic decision based on survival scores obtained from SELECTION. (c) Survival scores derived by SELECTION found in train, internal validation and external validation in each of the four cohorts in this study. Abbreviation: ATOM, AI-based Treatment Decision Model. P value shows two-sided Mann–Whitney test, Whiskers show minimum and maximum values, boxes represent 25–75% data ranges, ns denotes no significant differences (P > 0.05), ∗ denotes a P value < 0.05. SELECTION: unreSEctabLe hEpatocellular Carcinoma mulTImOdal transformer. DLS, Deep learning score representing the survival score computed by SELECTION in predicting ORR or OS; MSS, median survival score found in the training set; HCC, Hepatocellular Carcinoma; Train, Training datasets; Int, Internal Validation datasets; Ext, External Validation datasets. T-C, TACE with systemic treatment; H-C, HAIC with systemic treatment.

Fig. 4

Kaplan Meier overall survival analysis of high and low-risk groups for (a) TACE group; (b) HAIC group; (c) TC group; (d) HC group. Abbreviation: T-C, TACE with systemic treatment; H-C, HAIC with systemic treatment; HR, hazard ratio.

Fig. 5

Feature visualization of SELECTION models on four representative patients using Grad-Cam. Abbreviation: T-C, TACE with systemic treatment; H-C, HAIC with systemic treatment.